Rosie version 1.0.0-alpha-2 has been released, and the Python module is back!

There is a fresh implementation of, making Rosie available (again) from a variety of programming languages. A module for one language, Python, is part of the 1.0.0-alpha-2 release.

I am very interested in feedback and contributions to make better. In particular, more arguments can be made into keywords or given default values.

Also, the current api provided by returns the match data in a buffer created by the cffi package. This buffer does not have all the features of “standard” Python buffers or memory views. The api returns this buffer in order to avoid making a decision for the user about how the match data will be used. In other words, converting it to a string would require making a copy of the data, whereas often the buffer can be used directly instead.

Rosie can return match data in many forms, though probably the most useful for a Python programmer is the JSON format, which can be processed by json.loads to produce a Python structure.

My experience with JSON libraries leads me to think that many inefficiencies lurk there, however. This is why Rosie has its own JSON generator, tailored to implement only the needed features, and written in C

If Python’s json.loads is too slow, there is an attractive alternative. Rosie can produce a byte-encoded output that has a simple format. Match data in this compact encoding can be returned to Python, where it could be decoded in If anyone is interested in coding this up, let me know!


Another alpha release, coming soon, will include rich support for creating character classes based on Unicode properties.

Rosie modules for other languages, like Go, C, node.js, and Ruby, are also forthcoming. We had these for Rosie v0.99, and they are on the roadmap to reproduce for the new librosie in v1.0.0.

And will be revised, ideally with some help from Python programmers, to be more Pythonic.

The Rosie macro facility has been “dark launched” in the sense that the following macros are already part of Rosie v1.0.0: * find * findall * ci

Although macro documentation is lacking, some examples are lurking about. Here’s a teaser: The find macro skips ahead until its pattern argument is matched, and captures the pattern argument. Macros are invoked using the macro name followed by a colon. Here’s an example where we want to search for the word “nameserver” at the start of a line that also contains “201”:

$ rosie -o line match '"nameserver" find:"201"' test/resolv.conf 

Finally, a number of customization capabilities are being designed, along the lines of a ~/.rosierc file, for users of the Rosie CLI and REPL.

Whether you code in Python or any other language that has libffi support, you have access to #modernpatternmatching.

Follow us on Twitter for announcements. We expect v1.0.0-beta to be released before the end of 2017.